Flash

The news, the context, and what changes.

RSS
Flash

ByteDance Open-Sources Bernini, a Unified Video Generation and Editing Framework

ByteDance has officially open-sourced the Bernini framework, designed to unify video generation and editing workflows. The framework innovatively employs a multimodal large model as its planning core, which precisely understands user intent and semantic goals to guide diffusion models in producing high-quality visual renders, thereby significantly enhancing the controllability of AI video editing.

88Z AI editorial1 min read
Flash

Alibaba Tongyi Launches Multimodal Agent Foundation Model Qwen3.7-Plus

Alibaba Tongyi has launched Qwen3.7-Plus, a multimodal agent foundation model that unifies visual and language capabilities to support scene perception, GUI operation, and code generation. The model ranks fifth globally and first in China on the Vision Arena leaderboard, demonstrating the ability to handle complex workflows such as full-link app development.

88Z AI editorial1 min read
Flash

Peking University Team Upgrades ChatExcel Max 2.0

The Peking University team launched ChatExcel Max 2.0, removing file limits while preserving original formats. It supports conversational data processing and multi-format report generation, enabling automated analysis and data security for finance, logistics, and other sectors.

88Z AI editorial1 min read
Flash

Alibaba DAMO Academy Launches AI-Native Development Platform 'Yugou'

Alibaba DAMO Academy has launched Yugou, an AI-native development platform centered on Vibe Coding that enables users to generate complete digital products via natural language. The platform covers the entire workflow from requirement understanding to release and features built-in standardized resource interfaces for zero-code access.

88Z AI editorial1 min read
Flash

Alphabet Announces $80 Billion Financing for AI Spending

Google parent company Alphabet has announced an $80 billion equity financing round, primarily directed toward AI infrastructure and computing power. The funding comes from public offerings, at-the-market issuances, and a private investment by Berkshire Hathaway. This move aims to support the company's growing capital expenditure plans and strong cloud business demand.

88Z AI editorial1 min read
Flash

Ciyuan Wuxian Completes Tens of Millions in Financing

Ciyuan Wuxian has completed a tens-of-millions angel+ round of financing, co-led by Huakong Fund and Shuimu Ventures, with Xiajin Innovation participating. The founding team combines expertise from ByteDance's AI transformation projects and Tsinghua University's Yao Class. Focusing on enterprise-grade Agent infrastructure, its InfCode and InfOne products already serve leading financial and software companies.

88Z AI editorial1 min read
Flash

Coze 3.0 Officially Launched

Coze 3.0 has officially launched, covering iOS, Android, Mac, Windows, and web platforms. The new version focuses on AI team collaboration, supporting multi-Agent collaboration and project space management. It allows access to local Agents like Claude Code, provides long-term cloud running capabilities, and includes various professional templates.

88Z AI editorial1 min read
Flash

MiniMax Open-Sources New AI Model MiniMax M3

MiniMax has released the open-source MiniMax M3, the first domestic model combining advanced coding, million-token context, and native multimodal capabilities. Utilizing a new MSA sparse attention architecture, it achieves strong performance in international benchmarks, surpassing some leading overseas models.

88Z AI editorial1 min read
Flash

HaiSnap App Officially Launches

The HaiSnap app has officially launched as a mobile AI creation agent platform. Users can initiate tasks via voice, text, or file uploads, with AI automatically generating publishable outputs such as web apps, documents, PPTs, and images. Core features include an agent workspace, results library, task playback, and skill invocation.

88Z AI editorial1 min read
Flash

Xiaomi Open-Sources Controllable Video Sound Effect Generation Model ControlFoley

Xiaomi's large model application team has released the open-source model ControlFoley, which uniformly supports three video dubbing tasks: text-guided, text-controlled, and reference audio-controlled. By leveraging visual encoding and modal robust training, the model addresses issues of weak control and synchronization difficulties, achieving state-of-the-art performance among open-source solutions on multiple benchmarks.

88Z AI editorial1 min read